基于多判据神经网络的单相自适应重合闸的研究

Study on Single-phase Adaptive Auto-reclosure Based on Multi-criterion Neural Network

提出了一种基于多判据神经网络的电力系统单相自适应重合闸优化方案,该方法能够正确进行瞬时故障和永久故障的区分。当瞬时性故障发生时,在短路点电弧熄灭后的恢复电压阶段,断开相各电气量的关系与永久性故障将有本质的不同。在详细分析断开相工频电气量的基础上,用滤波后的采样值通过预处理层构成3种判据,利用神经网络将它们的自适应赋予权值,最后得出正确的结果。经过大量的仿真试验,该方案获得了满意的效果。

An optimized scheme based on multi-criterion neural network for single-pole adaptive reclosing in power system is proposed. It can distinguish transient fault from permanent fault occurred on the EHV transmission line properly. During the voltage recovery period after the extinction of fault are, the amplitude and phase of transient fault voltage is essentially different from those of permanent fault voltage. Based on the analysis of fault voltage using filtered sample data, three criterions are set up through per-processing layer, they are vested specific gravity adaptively through neural network, finally get the correct result. All the conclusions have been verified by simulation.